Agentic AI Practice 🚀

A hands-on collection of agentic applications built with LangChain, local LLMs (Ollama), and practical tool integration.

🎯 What's Here?

Each project demonstrates a specific pattern or capability in building autonomous agents:

• Tool calling & reasoning
• Multi-agent collaboration
• Memory & state management
• Custom tools & integrations
• Production patterns

📁 Current Projects

Project	Description	Guide
1_calculator_agent	ReAct agent with math calculator tool	Guide
1_calculator_agent_v2	Unified model-agnostic calculator agent — fallback default + any model/server via CLI flags	Guide V2

🛠️ Getting Started

# Clone and setup
python3 -m venv --system-site-packages .venv && source .venv/bin/activate
python3 -m pip install -r requirements.txt

If the pip/pip3 wrappers in .venv/bin fail with a "bad interpreter" error, the venv was created elsewhere and renamed. Recreate it with python3 -m venv --system-site-packages .venv. Using --system-site-packages inherits packages already installed system-wide, so you may not need to install anything.

🔧 Common Setup

All projects use local Ollama by default. Configure via environment variables or CLI flags:

# Env vars
export OLLAMA_BASE_URL="http://localhost:11434"
export OLLAMA_HOST="localhost"   # or 192.168.x.x for LAN
export OLLAMA_PORT="11434"

# Or CLI flags on 1_calculator_agent_v2.py (defaults: host=localhost, port=11434)
.venv/bin/python3 1_calculator_agent_v2.py --host 192.168.0.73 --port 11434 --model ministral-3:3b

📚 Learn More

• LangChain Docs
• Ollama Python Client
• ReAct Paper